CN110940882A - Electric energy quality disturbance identification method for optimizing S transformation by genetic algorithm - Google Patents
Electric energy quality disturbance identification method for optimizing S transformation by genetic algorithm Download PDFInfo
- Publication number
- CN110940882A CN110940882A CN201911263073.9A CN201911263073A CN110940882A CN 110940882 A CN110940882 A CN 110940882A CN 201911263073 A CN201911263073 A CN 201911263073A CN 110940882 A CN110940882 A CN 110940882A
- Authority
- CN
- China
- Prior art keywords
- transformation
- genetic algorithm
- transform
- disturbance
- generalized
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Abstract
The invention relates to a power quality disturbance identification method for genetic algorithm optimization S transformation, which comprises the following steps: acquiring a disturbance signal; optimizing the generalized S transformation through a genetic algorithm, and carrying out transformation processing on the disturbance signal through the optimized generalized S transformation to obtain a transformation result; and extracting a feature vector from the transformation result, and obtaining an identification result based on a support vector machine. Compared with the prior art, the method has the advantages of improving the feature extraction precision, realizing better disturbance classification effect and the like.
Description
Technical Field
The invention relates to the field of electric energy quality disturbance identification and classification, in particular to an electric energy quality disturbance identification method for optimizing S transformation by a genetic algorithm.
Background
With the continuous improvement of the living standard of residents, more and more nonlinear loads are connected into the existing power system, so that the quality of electric energy has a lot of problems. For optimizing and managing the power quality, firstly, detecting which types of disturbances exist in the power quality.
Since the classification and identification of the power quality disturbance is not a single disturbance but various changes and composite disturbances exist, various factors need to be considered in the process of feature extraction and identification, and therefore, the research on the power quality needs to be divided into different steps to be carried out: feature extraction and signal classification. The extraction of stable and effective characteristics is the key point in the rapid identification and research of the power quality. The international common concentrated feature extraction method comprises the following steps: fourier transform; short-time Fourier transform; performing wavelet transformation; dp changing; Hilbert-Huang transform; s transform, etc. The S Transformation (ST) is a time-frequency analysis method proposed by StockWell and the like, combines the advantages of Continuous Wavelet Transformation (CWT) and short-time Fourier transformation (STFT), has good time-frequency analysis capability, is particularly suitable for processing nonstationary signals such as electric energy quality disturbance signals, and is insensitive to noise. However, in practical applications in some fields, there are disadvantages that the time-frequency energy is not gathered when the S transform is used to perform time-frequency analysis on the signal. The power quality disturbance is a complex signal with multiple disturbance compounds and multiple noises superposed, and the problems of poor frequency spectrum resolution, unobvious characteristics or loss can be caused by directly carrying out S transformation on the complex signal to extract the characteristics.
However, in real-world applications, there are few cases of optimizing the window width factor or the cost function, and although there are many documents that optimize the window width factor or the cost function, satisfactory results cannot be achieved.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provide the electric energy quality disturbance identification method for optimizing the S transformation by the genetic algorithm, which improves the feature extraction precision and realizes a better disturbance classification effect.
The purpose of the invention can be realized by the following technical scheme:
a power quality disturbance identification method for genetic algorithm optimization S transformation comprises the following steps:
acquiring a disturbance signal;
optimizing the generalized S transformation through a genetic algorithm, and carrying out transformation processing on the disturbance signal through the optimized generalized S transformation to obtain a transformation result;
and extracting a feature vector from the transformation result, and obtaining an identification result based on a support vector machine.
Further, the generalized S transform expression is:
where f is frequency, m is control parameter, x (t) is input signal, t is time, and τ is the center point of the window function.
Further, the optimizing the generalized S transform by the genetic algorithm specifically includes:
the control parameters used in the generalized S-transform are optimized using a genetic algorithm.
Further, when the control parameters used in the generalized S transformation are optimized by using the genetic algorithm, the time-frequency distribution concentration is used as the evaluation standard of the optimal control parameters.
Further, the optimizing the control parameter used in the generalized S transform using the genetic algorithm specifically includes:
1) coding the control parameters to construct an initial population;
2) evaluating each individual in the current population by taking the time-frequency distribution concentration as the fitness to obtain a control parameter m under the current optimal fitnessbestAnd the current optimal individual;
3) performing crossover and mutation operations in sequence to update the population;
4) and judging whether the termination condition is met, if so, obtaining the final optimal individual and the corresponding optimal control parameter, and if not, returning to the step 2).
Further, the expression of the time-frequency distribution concentration ratio is as follows:
in the formula, Lx(m) is the time-frequency distribution concentration, S (N, k) is a function related to disturbance signals and control parameters, q is the current optimal individual, and N is the population number.
Further, the termination condition is that the fitness changes by < 1% or the number of iterations >100 times over 20 generations.
Further, in the initial population, each individual is composed of
Forming an n-dimensional vector.
Further, in the step 2), the smaller the concentration of time-frequency distribution, the better the control parameter.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention adopts the constructed generalized S transformation to carry out the electric energy quality disturbance identification, thereby solving the problem of non-centralized time-frequency energy;
2. the optimal control parameters are selected in a self-adaptive mode by utilizing a genetic algorithm, the problem that the standard difference values of all frequency bands are not changed along with the frequency change trend is solved, and the adaptability of S transformation in signal analysis is improved;
3. the invention optimizes the generalized S transformation by using the genetic algorithm, obviously eliminates most high-frequency useless noise, leads the energy to be more concentrated and smooth and can obviously extract the frequency.
4. The signal time-frequency energy processed by the optimized S transformation is more concentrated, and the invention can distinguish different types of electric energy quality disturbance.
Drawings
FIG. 1 is a process diagram of a disturbance identification method of the present invention;
FIG. 2 is a schematic diagram of the process of optimizing S-transform by genetic algorithm of the present invention;
FIG. 3 is a comparison graph of the transformation effect of a harmonic signal S, GA-S in an embodiment, wherein (3a) is a three-dimensional graph of S transformation, (3b) is a three-dimensional graph of GA-S transformation, (3c) is a frequency-amplitude graph of S transformation, and (3d) is a frequency-amplitude graph of GA-S transformation;
fig. 4 is a graph comparing the transformation effect of a transient oscillation signal S, GA-S in an embodiment, wherein (4a) is a three-dimensional graph of S transformation, (4b) is a three-dimensional graph of GA-S transformation, (4c) is a frequency-amplitude graph of S transformation, and (4d) is a frequency-amplitude graph of GA-S transformation.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
As shown in fig. 1, the embodiment provides an electric energy quality disturbance identification method for genetic algorithm optimization S transformation, which optimizes generalized S transformation through a genetic algorithm, transforms the disturbance signal through the optimized generalized S transformation, extracts a feature vector from a transformation result, and obtains an identification result based on a support vector machine. The method adopts an improved S transformation method to perform self-adaptive feature extraction on the power quality disturbance signal, and meanwhile, the minimum time-frequency distribution concentration value is taken as a window control parameter evaluation standard to realize self-adaptive selection of appropriate control parameters to extract the most effective features so as to realize a better disturbance classification effect.
a. Generation of power quality disturbance signals
The transformation formula of S is:
where f is the frequency, x (t) is the input signal, t is the time, τ is the center point of the window function, ω (t- τ) is the Gaussian window, and is expressed as:
and also
Therefore, S (. tau.f) can be rewritten as shown in formula (4):
as can be seen from equation (3), the advantage of S transformation is that the standard deviation σ is a function of the frequency f, so the height and width of the gaussian window can vary with the frequency, but all frequency bands of the S transformation are provided with constant standard deviation, and the trend of the standard deviation value varying with the frequency is constant, which affects the adaptability of S transformation in the power quality signal processing.
The method adopts generalized S transformation to solve the problem of poor S transformation adaptability, and the formula is as follows:
where f is frequency, m is control parameter, x (t) is input signal, t is time, and τ is the center point of the window function.
When m is 1, formula (5) is the standard S transform; as can be seen from equation (5), the generalized S transformation improves σ to be
The size of the standard deviation is changed through a given control parameter m, so that the width of a window width function is controlled, and the time-frequency energy aggregation performance is better. In general, m is in the range of (0, 1)]. The method provides a method for optimizing GST (generalized S transform) by using genetic calculation to search an optimal m value; and setting the time-frequency distribution concentration Lx(m) is an optimal m-value evaluation standard, and the standard is LxThe smaller the value of (m), the more concentrated the energy distribution, which also represents the optimal value of m. The time-frequency distribution concentration is as follows:
in the formula, Lx(m) is the time-frequency distribution concentration,
and after normalization, the number is 1, representing unbiased energy, q is the current optimal individual, and N is the population number.
Disturbance noise in actual power quality disturbance signals is too much, but most common disturbances are only limited to harmonic disturbance and cannot completely reflect the diversity of the disturbance signals. The embodiment randomly generates 210 training samples and 1120 testing samples by using the power quality disturbance model, wherein the number of the training samples and the number of the testing samples in each class are respectively 30 and 160. And Gaussian white noise with the signal-to-noise ratio of 30-50 db is randomly added into all the signals, and the sampling frequency is set to be 3.2 KHz.
b. GA-S conversion of original disturbance signal
The GA optimization GST procedure was as follows:
(a) taking the parameter m as a chromosome, randomly generating N binary coded chromosomes and coding the m value to form an initial population
Wherein
Represents the jth individual of the a-th generation.
n is the length of one chromosome.
(b) And evaluating the fitness of the current population. Initialization control parameter selection range (0, 1)]. Converting randomly generated parameters to be selected into decimal values to form a parameter vector with N elements, substituting each value into generalized S transformation, respectively processing various power quality disturbances, and carrying out fitness (time-frequency distribution concentration q)best) Calculating to maintain the current best fitness (L)xControl parameter m at (m) value minimum)bestAnd the current generation of the optimal individual qbest。
(c) A new population S (1) is obtained through cross variation updating, the cross rate is 0.5, and the variation rate is 0.8.
(d) Judging whether a termination condition is met, and stopping if the termination condition is met; otherwise, let a be a +1, return to step (b). Continuously iterating and updating the process to finally obtain the optimal control parameter m of the generalized S transformationout。
Before the genetic algorithm optimizes the generalized S change control parameters, initial values need to be set for some parameters in the algorithm, and the parameters are set as follows;
set to each α in S (0)
It means that all states of a chromosome are superposed with the same probability at the time of initialization, the length N of the chromosome is 20, and the number N of the chromosomes is 20, namely the population size. The maximum iteration number maxgen is 40, and the algorithm stop condition is that the fitness changes more than 20 generations<1% or number of iterations>The treatment is carried out 100 times.
In view of the fact that the number of all the transformed comparison pictures is large, in the embodiment, two typical waveforms are selected, and two effect graphs obtained after different transformations are adopted for comparison and evaluation (in order to verify the engineering application capability of the optimization algorithm, gaussian white noise with a signal-to-noise ratio of 50db is added into the two signals).
It can be seen from the harmonic transformation effect graph in fig. 3 that the graph energy after GA-S transformation is more concentrated and smooth, the frequency and amplitude characteristics can be obviously extracted, fundamental waves 50Hz, 5 th harmonic and 7 th harmonic are respectively added to the simulation signal, the GA-S frequency-amplitude graph can obviously extract the corresponding frequency and amplitude characteristics, and compared with the S transformation, not only all harmonics can not be identified, but also the amplitude information is very fuzzy.
From the transient oscillation transformation effect diagram of fig. 4, it can be seen that the GA-S obviously eliminates most of the high-frequency unwanted noise, the frequency characteristic is very obvious, the 50Hz power frequency is very obvious compared with the S transformation effect, and there is no aliasing phenomenon at the high frequency. The same GA-S time-amplitude surface effect is superior to S transformation, amplitude fluctuation can be obviously seen, and S transformation can not see fluctuation and has amplitude tearing phenomenon.
c. Extracting elements required for classifying feature vectors
Respectively extracting features from original disturbance signal data and a GA-S operation result matrix, and totally extracting 8 features for establishing an SVM classification tree, wherein the features are respectively as follows:
f1: averaging the amplitude values over time;
f2: standard deviation of amplitude within the fundamental;
f3: an amplitude less than 95%;
f4: an amount of amplitude greater than 105%;
f5: an amplitude less than 10%;
f6: the frequency corresponding to the second maximum point of the frequency;
f7: the frequency corresponding to the third maximum point of the frequency;
f8: frequency standard deviation square root mean.
d. Structure of SVM classification tree of design
The classification disturbance types are 7 types in total, and are 4 types of single disturbance, namely voltage temporary rise, voltage temporary drop, voltage interruption and transient oscillation respectively; and three composite disturbances of harmonic wave + pause, harmonic wave + pause and harmonic wave + interruption are also included, so that 6 SVM classifiers are arranged to form a classification tree.
e. The effectiveness of the invention was verified using simulated signals and compared to the unmodified method
And respectively constructing a GA-S support vector machine classification tree and a traditional ST support vector machine classification tree by using 1330 groups of disturbance signals randomly generated by MATLAB, wherein the overall classification effect of the GA-S support vector machine classification tree is more than 98%. The classification results are compared to reflect the effectiveness of the improved algorithm on performance improvement, and the developed comparison results are shown in tables 1 and 2.
TABLE 1 Classification accuracy comparison for cases where the signal-to-noise ratio is 30-50 dB random values
As can be seen from Table 1, the method of the present invention improves the classification accuracy greatly.
Table 2 comparison of classification accuracy for cases with 30dB, 40dB, 50dB signal-to-noise ratio, respectively.
As can be seen from Table 2, the total classification accuracy of the method of the invention under different noise levels is higher than that of the method before improvement, and is kept above 97.8%, so that the method of the invention has good noise immunity and robustness.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions that can be obtained by a person skilled in the art through logic analysis, reasoning or limited experiments based on the prior art according to the concept of the present invention should be within the protection scope determined by the present invention.
Claims (9)
1. A power quality disturbance identification method for genetic algorithm optimization S transformation is characterized by comprising the following steps:
acquiring a disturbance signal;
optimizing the generalized S transformation through a genetic algorithm, and carrying out transformation processing on the disturbance signal through the optimized generalized S transformation to obtain a transformation result;
and extracting a feature vector from the transformation result, and obtaining an identification result based on a support vector machine.
2. The method for identifying power quality disturbance through genetic algorithm optimization S transform as claimed in claim 1, wherein the generalized S transform expression is:
where f is frequency, m is control parameter, x (t) is input signal, t is time, and τ is the center point of the window function.
3. The electric energy quality disturbance identification method for genetic algorithm optimization S transformation according to claim 1, wherein the optimization of the generalized S transformation through the genetic algorithm is specifically:
the control parameters used in the generalized S-transform are optimized using a genetic algorithm.
4. The method for identifying the power quality disturbance of the genetic algorithm optimized S transform as claimed in claim 3, wherein the control parameters used in the generalized S transform are optimized by the genetic algorithm, and the time-frequency distribution concentration is used as an evaluation criterion of the optimal control parameters.
5. The method for identifying the power quality disturbance of the genetic algorithm optimized S transform as claimed in claim 3, wherein the optimization of the control parameters used in the generalized S transform by the genetic algorithm specifically comprises the following steps:
1) coding the control parameters to construct an initial population;
2) evaluating each individual in the current population by taking the time-frequency distribution concentration as the fitness to obtain a control parameter m under the current optimal fitnessbestAnd the current optimal individual;
3) performing crossover and mutation operations in sequence to update the population;
4) and judging whether the termination condition is met, if so, obtaining the final optimal individual and the corresponding optimal control parameter, and if not, returning to the step 2).
6. The method for identifying the power quality disturbance of the genetic algorithm optimized S transform as claimed in claim 5, wherein the expression of the time-frequency distribution concentration is as follows:
in the formula, Lx(m) is the time-frequency distribution concentration, S (N, k) is a function related to disturbance signals and control parameters, q is the current optimal individual, and N is the population number.
7. The method for identifying power quality disturbance through genetic algorithm optimization S transformation as claimed in claim 5, wherein the termination condition is that the fitness changes by < 1% or the iteration number >100 times for more than 20 generations.
8. The method for identifying power quality disturbance through genetic algorithm optimization S transform as claimed in claim 5, wherein each individual in the initial population is selected from
Forming an n-dimensional vector.
9. The method for recognizing the power quality disturbance of the genetic algorithm optimized S transform as claimed in claim 5, wherein in the step 2), the smaller the concentration of time-frequency distribution, the better the control parameters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911263073.9A CN110940882A (en) | 2019-12-11 | 2019-12-11 | Electric energy quality disturbance identification method for optimizing S transformation by genetic algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911263073.9A CN110940882A (en) | 2019-12-11 | 2019-12-11 | Electric energy quality disturbance identification method for optimizing S transformation by genetic algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110940882A true CN110940882A (en) | 2020-03-31 |
Family
ID=69910669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911263073.9A Pending CN110940882A (en) | 2019-12-11 | 2019-12-11 | Electric energy quality disturbance identification method for optimizing S transformation by genetic algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110940882A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2798189C1 (en) * | 2022-06-20 | 2023-06-16 | федеральное государственное бюджетное образовательное учреждение высшего образования "Ульяновский государственный технический университет" | Method for determining proportion of interference energy generated by an electrical consumer |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102608441A (en) * | 2011-01-25 | 2012-07-25 | 华北电力科学研究院有限责任公司 | S conversion based SDPQD (Short Duration Power Quality Disturbance) signal identifying method, device and system |
| CN104459397A (en) * | 2014-12-08 | 2015-03-25 | 东北电力大学 | Power quality disturbance recognizing method with self-adaptation multi-resolution generalized S conversion adopted |
-
2019
- 2019-12-11 CN CN201911263073.9A patent/CN110940882A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102608441A (en) * | 2011-01-25 | 2012-07-25 | 华北电力科学研究院有限责任公司 | S conversion based SDPQD (Short Duration Power Quality Disturbance) signal identifying method, device and system |
| CN104459397A (en) * | 2014-12-08 | 2015-03-25 | 东北电力大学 | Power quality disturbance recognizing method with self-adaptation multi-resolution generalized S conversion adopted |
Non-Patent Citations (3)
| Title |
|---|
| 杨剑锋等: ""基于分段改进S变换的复合电能质量扰动识别"", 《电力系统保护与控制》 * |
| 林琳等: ""基于遗传算法优化支持向量机的电能质量暂态扰动识别新方法"", 《水电能源科学》 * |
| 王波等: ""基于QGA 优化广义S 变换的滚动轴承故障特征提取"", 《振动与冲击》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2798189C1 (en) * | 2022-06-20 | 2023-06-16 | федеральное государственное бюджетное образовательное учреждение высшего образования "Ульяновский государственный технический университет" | Method for determining proportion of interference energy generated by an electrical consumer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mahela et al. | Recognition of power quality disturbances using S-transform based ruled decision tree and fuzzy C-means clustering classifiers | |
| Qiu et al. | Power quality disturbances recognition using modified s transform and parallel stack sparse auto-encoder | |
| Gunal et al. | The search for optimal feature set in power quality event classification | |
| CN110443302B (en) | Load identification method based on feature fusion and deep learning and application thereof | |
| Wan et al. | Pattern recognition of partial discharge image based on one-dimensional convolutional neural network | |
| Biswal et al. | Power quality disturbance classification using fuzzy C-means algorithm and adaptive particle swarm optimization | |
| CN107462785B (en) | The more disturbing signal classifying identification methods of power quality based on GA-SVM | |
| Karasu et al. | Investigation of power quality disturbances by using 2D discrete orthonormal S-transform, machine learning and multi-objective evolutionary algorithms | |
| CN106529428A (en) | Underwater target recognition method based on deep learning | |
| Xiao et al. | High-impedance fault detection method based on one-dimensional variational prototyping-encoder for distribution networks | |
| CN109919364A (en) | Multivariate Time Series prediction technique based on adaptive noise reduction and integrated LSTM | |
| CN109948726B (en) | Power quality disturbance classification method based on deep forest | |
| CN111476173A (en) | BAS-SVM-based power distribution network voltage sag source identification method | |
| CN109934303A (en) | A kind of non-invasive household electrical appliance load recognition methods, device and storage medium | |
| CN111368892A (en) | Generalized S transformation and SVM electric energy quality disturbance efficient identification method | |
| CN113408341A (en) | Load identification method and device, computer equipment and storage medium | |
| CN110276256A (en) | Based on the low signal-to-noise ratio Modulation Recognition of Communication Signal method and device for adjusting ginseng accidental resonance | |
| CN115758249A (en) | Non-intrusive load identification method and device | |
| CN113378652B (en) | Disturbance classification method based on EWT-MPE-PSO-BP | |
| CN112686182A (en) | Partial discharge mode identification method and terminal equipment | |
| CN107329473A (en) | The robot polling path planing method searched for based on average harmony | |
| CN110940882A (en) | Electric energy quality disturbance identification method for optimizing S transformation by genetic algorithm | |
| Manikonda et al. | Power quality event classification using transfer learning on images | |
| Liu et al. | Classification and identification of electric shock current for safety operation in power distribution network | |
| Li et al. | Emitter signals modulation recognition based on discriminative projection and collaborative representation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200331 |